av1_c_vs_simd_encode.sh: fix missing cpu-used vals for arm
fixes:
seq: missing operand
after:
e845a86a42 Modify c vs SIMD test
The result was no good quality encodes were being run for arm, just
realtime.
Change-Id: Ib90ae25d40daa16e4ec46c734b0d9a5d80b91945
diff --git a/test/av1_c_vs_simd_encode.sh b/test/av1_c_vs_simd_encode.sh
index 4fc481a..d2e4e9b 100755
--- a/test/av1_c_vs_simd_encode.sh
+++ b/test/av1_c_vs_simd_encode.sh
@@ -355,13 +355,23 @@
fi
if [ "${preset}" = "good" ]; then
- if [ "${arch}" = "x86_64" ]; then
- local min_cpu_used=0
- local max_cpu_used=6
- elif [ "${arch}" = "x86" ]; then
- local min_cpu_used=2
- local max_cpu_used=3
- fi
+ case "${arch}" in
+ arm64)
+ # Speed 0 is not tested as arm64 is run under emulation.
+ local min_cpu_used=1
+ local max_cpu_used=6
+ ;;
+ x86)
+ # x86 has a good amount of overlap with x86-64. Only a few values are
+ # tested to improve the runtime of the script.
+ local min_cpu_used=2
+ local max_cpu_used=3
+ ;;
+ *)
+ local min_cpu_used=0
+ local max_cpu_used=6
+ ;;
+ esac
local test_params=av1_encode_good_params
elif [ "${preset}" = "rt" ]; then
local min_cpu_used=5
